1. The stress gradient hypothesis (SGH) predicts a shift from net negative interactions in benign environments towards net positive in harsh environments in ecological communities. While several studies found support for the SGH, others found evidence against it, leading to a debate on how nature and strength of species interactions change along stress gradients, and to calls for new empirical and theoretical work.
2. In the latest attempt in this journal, it is successfully argued how the SGH should be expanded by considering different life strategies of species (stress tolerance versus competitive ability) and characteristics of abiotic stress (resource versus non-resource based) over wider stress gradients (opposed to low–high contrasts), but the crucial role of biotic stress by consumers is largely ignored in this refinement.
3. We point out that consumers strongly alter the outcome of species interactions in benign and harsh environments, and show how inclusion of consumer-incurred biotic stress alters the predicted outcome of interactions along resource- and non-resource-based stress gradients for stress-tolerant and competitive benefactors and beneficiaries.
4.Synthesis. New studies should include stress gradients consisting of both abiotic and biotic components to disentangle their impacts, and to improve our understanding of how species interactions change along environmental gradients.
In the latest attempt, published in this journal, Maestre et al. (2009) convincingly argue that consideration of the life history of interacting species (competitor versus stress tolerant, sensuGrime (1977)) and differentiating between types of stressors (resource versus non-resource based) is needed to refine the predictions of the SGH. The authors also rightly stress that sufficiently wide stress gradients need to be considered in new empirical studies, opposed to presently prevailing two-levelled contrasts (low versus high stress), which logically always result in linear relationships (see also Kawai & Tokeshi (2007), for similar argumentation). But where their attempt fails – and so far nearly every other recent study on the SGH – is that the crucial role of biotic stress incurred by consumers remains largely untouched. Consumer pressure, however, was an essential element in the original SGH models for both plant and animal communities, but it seems that consumers are overlooked in the present SGH studies, despite the evident role they play in communities in various terrestrial and aquatic environments (Cyr & Face 1993; Olff & Ritchie 1998).
The importance of biotic stress
There is abundant evidence that biotic stress by consumers significantly alters the nature and strength of species interactions in ecological communities and can even swap the sign of species interactions between benefactors and beneficiaries. For example, it is in the presence of consumers that consumer-defended species (physically or chemically defended by spines, alkaloids or high lignin content) positively affect the survival and growth of consumer-undefended species – a phenomenon also called associational resistance, associational avoidance, associational defence or defence guild – while in the absence of consumers the net outcome of this interaction is predominantly negative (Atstatt & Odowd 1976;Hay 1986; Rousset & Lepart 1999, 2000; Bakker et al. 2004; Callaway et al. 2005; Smit, den Ouden & Muller-Scharer 2006). Also, the related opposite process, associational palatability, which occurs when having tasty neighbours results in negative impact by increased consumption on consumer-defended species, is frequently observed in the presence of consumers (Atstatt & Odowd 1976; Rousset & Lepart 2002; Baraza, Zamora & Hódar 2006). Some of these studies on consumer-altered species interactions already date back to the time before the original SGH and give excellent examples of various plants, sessile invertebrates and consumers from both terrestrial and aquatic ecosystems, while the current SGH studies primarily focus on vascular plants in terrestrial ecosystems (without incorporating consumers).
Furthermore, recent studies show that increasing biotic stress results in a hump-shaped relationship rather than a linear relationship (Brooker et al. 2006; Smit et al. 2007; Vandenberghe et al. 2008; Levenbach 2009). This means that instead of a linear shift from net negative to net positive, as predicted by the original SGH, there might be an optimum in net positive interactions at intermediate stress levels. Explanations for this hump shape are that at low to intermediate stress levels the outcome of a benefactor–beneficiary interaction shifts from net negative to net positive, whereas at high stress levels both the performance (survival and growth) of the benefactor and the beneficiary declines, leading to the disappearance of the net positive effect. In the case of Levenbach (2009), low biotic stress by consumers (sea urchins) resulted in competition for space between benthic turf algae and the strawberry anemone Corynactis californica, and intermediate biotic stress resulted in facilitation by anemones for turf algae because the anemones physically protect themselves and associated algae via their nematocysts (venomous subcellular organelles also known from e.g. stinging jellyfish) against sea urchins. These positive effects disappeared at high levels of biotic stress, probably because increased food (algae) shortage forced the sea urchins to forage closer to the unsafe anemones, and the subsequent higher contact rates with the anemones may have reduced the anemones’ deterring mechanism protecting against sea urchin grazing. In the case of Smit et al. (2007) and Vandenberghe et al. (2008), low biotic stress by consumers (cattle grazing) resulted in net neutral effects of the spiny shrub Rosa rubiginosa on the performance (survival and height growth) of four species of tree seedlings, whereas intermediate stress resulted in net positive effects, because the spiny shrubs physically protected the associated beneficiary (tree seedling) from grazing. High biotic stress resulted in the destruction of the benefactor, and due to this diminished physical protection the survival and growth of the associated tree seedlings declined rapidly. These three studies so far are clear examples of the impact of increased biotic stress (consumer pressure) on the net outcome of species interactions, and it is likely that this is true for environments ranging from low to high abiotic stress (Smit et al. 2007). Indeed, described positive and negative species interactions induced by biotic stress have been reported under both low- and high-abiotic stress conditions, while recent work even claims that particularly these positive biotic stress-induced effects – rather than abiotic stress relief – should predominate under abiotic stressful conditions (Graff, Aguiar & Chaneton 2007).
Inclusion of biotic stress along abiotic stress gradients
Hence, studies on the SGH have so far primarily focussed on species interactions along abiotic stress gradients, often a combination of stress factors involving drought, salinity, temperature or water logging. The net outcome of these interactions is typically expressed as relative interaction indices or relative neighbour effects (RNE), presenting the performance of a beneficiary in the presence of a benefactor compared to controls without a benefactor (e.g. Armas, Ordiales & Pugnaire 2004; Maestre & Cortina 2004; Brooker et al. 2006). Both measures indicate the stress experienced by beneficiary plants in various environments in a standardized way. Yet, as mentioned before, the significant role of biotic stress incurred by consumers along the studied abiotic stress gradients is thus far generally ignored or not controlled for (but see Crain 2008;Eskelinen 2008). However, given that consumers and the biotic stress they impose play a crucial role in nearly all terrestrial or aquatic ecosystems, from benign to harsh environments, such biotic stress along different resource- and non-resource-based stress gradients should be incorporated in the SGH. We now provide examples of how such inclusion of biotic stress by consumers along abiotic stress gradients would alter the predictions from the refined SGH (Maestre et al. 2009) using RNE. For simplicity, we first assume that consumer pressure remains constant over the abiotic stress gradient in these examples.
To start with, adaptations of plants to abiotic stress (resource and non-resource based) may simultaneously work as protection against consumers. For instance, species adapted to drought may have tiny, thick, spiny leaves, often covered with thick layers of cuticula or waxes to reduce excessive evapotranspiration and protect against high solar radiation, and these adaptations generally also work as deterrents against consumers. Therefore, such stress-adapted benefactors may protect beneficiaries not only against abiotic stress (e.g. through shading), but also against biotic stress (consumers), leading to better performance of beneficiaries (e.g. survival, growth, reproduction) compared to controls growing without benefactors. However, when consumers are not present, this interaction may easily be neutral or negative compared to controls without benefactors, e.g. because of competition for nutrients with the benefactor. For instance, along a non-resource-based stress gradient (e.g. salinity), the RNE of a stress benefactor (e.g. salt-tolerant, unpalatable) on a competitive beneficiary (e.g. salt-intolerant, palatable) shifts from negative (−) to positive (+) to very positive (++) in the absence of consumers (Maestre et al. 2009), as the initial competition for nutrients with the benefactor at low stress is replaced by abiotic stress relief by the benefactor at high stress (e.g. reduced soil salinity by salt-tolerant benefactors for associated salt-intolerant beneficiaries (Bertness & Ewanchuk 2002)). In the presence of consumers, the RNE may be positive along the whole gradient, due to associational resistance at low stress and abiotic stress relief at high stress (Fig. 1a). We note here that we do not include the predictions of very positive effects (++) after death of the benefactor at high stress by Maestre et al. (2009) in our Fig. 1. Some positive effects of the dead benefactor on the beneficiary might indeed remain for a short time (microclimate or soil properties), but the main stress type experienced by beneficiaries with and without benefactor will become more similar, leading to a neutral or, at the most, a positive RNE, but not to a very positive RNE. When the benefactor is a competitor and the beneficiary is stress-tolerant (in either resource- or non-resource-based stress) and in absence of consumers, the refined SGH (Maestre et al. 2009) predicts that with increased stress (e.g. nutrient limitation) the RNE shifts from negative (−) to positive (+) to very positive (++). Indeed, compared to controls without benefactors, competition with benefactors for nutrients at low stress is outweighed by habitat amelioration at high stress (Callaway et al. 2002). In the presence of consumers, the RNE may remain negative as microhabitat amelioration is counterbalanced by associational palatability: consumers are attracted towards palatable benefactors leading to a disadvantage for associated beneficiaries compared to controls without benefactors (Fig. 1b). Lastly, when both benefactors and beneficiaries are competitors, along a resource-based stress gradient (e.g. nutrient limitation) and in absence of consumers, the RNE may go from neutral at low nutrient limitation to negative at medium and high nutrient limitation compared to controls, due to increased competition for resources (Tilman 1982). With consumers, the RNE may be positive at low nutrient limitation, as benefactors can use abundant available nitrogen compounds to produce alkaloids for their own and their associated beneficiaries’ defence. Such alkaloid-based defences are a common phenomenon in several species (e.g. Ranunculaceae, Solanaceae, Fabaceae, Taxus) particularly when sufficient nutrients are available, while the production of these alkaloids declines with nutrient limitation (Coley, Bryant & Chapin Iii 1985). So, the RNE may be reduced to neutral and negative at medium to high nutrient limitation due to reduced alkaloid production in the benefactor (Fig. 1c). We note here that also for this last example the outcome in absence of consumers opposes the predicted outcome of the example of Maestre et al. (2009); hence, we question the generality of the predictions of the refined SGH.
In our examples, we showed how inclusion of biotic stress incurred by consumers can significantly alter the predictions from the refined SGH (Maestre et al. 2009). For simplicity, we kept consumer pressure constant over the stress gradient in these examples. This can simply be applied in experimental or modelling studies and offers good opportunities to test hypotheses as given in Fig. 1. A big challenge may lie in assessing the outcome of interactions along resource- and non-resource-based stress gradients under natural occurring biotic stress. As other studies have suggested, it is likely that biotic and abiotic stress factors strongly interact (Crain 2008; Eskelinen 2008), which ultimately affects the outcome of species interactions. Thus far, general predictions of the outcome of species interactions along such combined stress gradients are difficult to make, as they depend on the intensity and combination of abiotic stress type (resource versus non-resource based), biotic stress type (consumer pressure, consumer type (selectivity of feeding behaviour)) and the life history of interacting species. Nevertheless, we would like to cautiously propose two contrasting hypotheses on how species interactions might change along gradients of combined abiotic and biotic stress, the first based on the idea that effects of consumers decrease with abiotic stress, the second based on the idea that consumers are most important at intermediate abiotic stress. The reasoning behind the first idea – benign environments should be able to support more consumers than harsh environments – is that species in harsh environments have developed stress tolerance strategies and grow slowly (Grime 1977) offering very little nutritional value to support consumers, in contrast to benign environments that contain fast-growing high-quality food for many consumers. Studies of both terrestrial and aquatic systems support this view (McNaughton et al. 1989; Cyr & Face 1993; Cebrian & Duarte 1994). Following from this, we may expect that associational resistance, i.e. benefactors protecting beneficiaries against consumers, is of little or no importance at high abiotic stress but increases in importance with decreasing abiotic stress (hypothesis 1). This first hypothesis is quite similar to predictions by Bruno, Stachowicz & Bertness (2003) in their altered Menge–Sutherland model (Menge & Sutherland 1976) after inclusion of facilitation. Our second alternative hypothesis is based on the ‘exploitation ecosystem hypothesis’ (Oksanen et al. 1981; Oksanen 1983) stating that consumers are particularly important at intermediate harsh–benign environments. Benign environments can support high quantities of consumers, but these consumers, and their effects, are heavily controlled by predators (top-down controlled). Harsh environments, on the other hand, hardly support any consumers, due to stress-adapted species offering low-quality food (similar argumentation as before). Several studies from both terrestrial and aquatic systems found supporting evidence for this (Virtanen 2000; Graff, Aguiar & Chaneton 2007; van der Wal & Hessen 2009). Following from this, we may expect that the importance of associational resistance shows a hump-shaped curve with increasing abiotic stress (hypothesis 2). Thus far, the here postulated alternative hypotheses have not been tested, but experimental studies covering sufficiently large environmental gradients to capture both extremes of the stress types in order to allow generalizations seem quite feasible for both terrestrial and aquatic systems. Hence, ultimately, new studies should include sufficiently wide stress gradients to avoid idiosyncrasies and scale effects, with gradients consisting of both abiotic and biotic components in order to disentangle the different impacts on species interactions for further refinement of the SGH. We particularly advocate the use of making and validating mechanistic models when formulating and testing specific hypotheses of how species interactions change along stress gradients. We look forward to these innovative contributions.